This turns reference images into parametric ForgeCAD models by treating photos as evidence about a real object, not targets to match pixel by pixel. You build the full 3D geometry first (including hidden sides you have to infer), validate it from canonical views, then calibrate cameras to compare against the originals. The workflow is rigid: write a Real Object Brief before modeling, blockout major volumes before chasing details, and test every reference image as a constraint instead of picking one hero shot. It includes a Python comparison board script and leans hard on the idea that if your model only looks right from the original camera angle, you built a facade instead of an object. Use it when you need actual geometry that holds up under inspection, not a pretty render that falls apart when rotated.
npx -y skills add kostard/forgecad-public-kit --skill forgecad-image-replicator --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
sickn33/antigravity-awesome-skills
kubesphere/kubesphere
supercent-io/skills-template